Content Calendar Tool

A complete example of an AI-powered content calendar plugin that plans editorial calendars, researches topics, and drafts channel-optimized content.

What It Does

The Content Calendar plugin:

1. Stores a user's content strategy profile (business description, audience, goals, channels, keywords)
2. Uses context.search() to research trends, competitors, and search intent
3. Generates a dated editorial calendar with AI using context.ai.analyze()
4. Drafts channel-optimized content for LinkedIn, X/Twitter, and SEO blog posts
5. Provides a launchable widget for browsing and managing the calendar

Tool Contract

The plugin exposes one tool — content_calendar — with these actions:

Action	Description	Key Parameters
show	Opens the calendar widget	—
setup_strategy	Stores content strategy	businessSummary, audience, offers, goals, preferredChannels
plan_calendar	Generates calendar items	startDate, endDate, instructions?
list_items	Lists calendar items	channel?, status?, startDate?, endDate?
items_for_date	Items for a specific date	date (YYYY-MM-DD or "tomorrow")
update_item	Edits a calendar item	itemId, plus any fields to update
refresh_research	Re-runs web search	itemId?, query?
write_item	Drafts one calendar item	itemId
write_for_date	Drafts all ready items for a date	date

Database Schema

The plugin uses four single-user tables with pie_user_id RLS:

• cc_strategy_profiles — One row per user. Business summary, audience, offers, goals, voice, channels, cadence, keywords.
• cc_research_runs — One row per search pass. Query, provider, normalized results, summaries, takeaways.
• cc_calendar_items — One row per content piece. Date, channel, status, pillar, title, hook, angle, CTA, brief, SEO fields.
• cc_drafts — One row per draft version. Item ID, version number, content payload, model metadata.

Status Flow

idea → scheduled → ready_to_write → drafting → drafted
                                   ↘ needs_refresh ↗
                                      archived

Using context.search() for Research

The plugin uses the platform search primitive to gather topical context:

var searchResult = await context.search('b2b saas content marketing trends 2026', {
  type: 'web',
  limit: 8,
  includeText: true,
  includeHighlights: true,
});

// Normalize results into research context
for (var i = 0; i < searchResult.results.length; i++) {
  var r = searchResult.results[i];
  console.log(r.title, r.url, r.snippet);
}

Research results are stored in cc_research_runs and linked to calendar items via research_run_id.

Prompt Stack Architecture

The plugin uses a layered prompt builder for AI generation:

1. Base system prompt — Action-specific instructions (plan_calendar, write_item, etc.)
2. Content Playbook — Verbatim injection of the full content writing playbook with forbidden lexicon, formatting rules, and platform-specific optimization guides
3. Channel guide — LinkedIn, X/Twitter, or blog-specific writing rules
4. User strategy — Business summary, audience, offers, goals, voice
5. Research context — Normalized search findings, source URLs, takeaways
6. Calendar item brief — Title, angle, hook, CTA, date, channel, SEO fields
7. Output schema — JSON structure expected in the response

Model Policy

All core AI tasks use google/gemini-3.1-pro-preview:

var result = await context.ai.analyze({
  prompt: fullPromptStack,
  data: { startDate: '2026-04-01', endDate: '2026-04-30', channels: ['linkedin', 'twitter', 'blog'] },
  model: 'google/gemini-3.1-pro-preview',
});

Cross-Chat Drafting

Any PIE chat can invoke the calendar tool to draft content:

"Write the content from the calendar for tomorrow"

This maps to the write_for_date action, which:

1. Resolves "tomorrow" to a concrete date
2. Finds all items with status scheduled, ready_to_write, or needs_refresh
3. Generates drafts for each item using the full prompt stack
4. Stores each draft as a versioned row in cc_drafts
5. Updates each item's status to drafted

Widget UI

The launchable widget provides:

• Calendar view — Month grid with channel-colored dots per day
• Agenda view — Chronological list grouped by date with status badges
• Strategy form — Editable profile for business gist, audience, channels, and keywords
• Plan modal — Date range picker with optional instructions for AI calendar generation
• Item detail — Full brief display with write/research/delete actions and draft preview

The widget communicates with the handler via PIE.sendAction() and receives updates via PIE.onData(). Dark mode is supported via PIE.onTheme().